Trust & data

Nerve reads your work to brief you, and acts only with your approval. To do that job it needs the access a human chief of staff would have. Here is exactly what it can touch, what it does with the data, and the parts we are still working on. No hiding the uncomfortable bits.

What Nerve accesses, and why

Gmail / Outlook
Reads: Your mail, to build your morning briefing and track what is slipping.
Can change: Sends, replies, drafts, and labels, but only the specific action you reviewed and approved first.
Google / Microsoft Calendar
Reads: Your events, to prep you and catch conflicts.
Can change: Creates or moves an event only when you approve that specific action.
Slack
Reads: The channels you add Nerve to, to catch what matters.
Can change: Posts only the message you approved.
Google Drive
Reads: Read-only, for documents you point it at.
Can change: Nothing. Read-only.

Today, connecting asks for send and edit permissions up front so an action you approve can execute without a second prompt. Every send and every edit is still gated behind your explicit review of that specific action. We are moving to incremental permissions: read-only to build your briefing, with write access requested only when you approve your first outbound action.

What Nerve never does

  • Never acts without your approval. Agents draft and queue. You approve or edit. Only then does anything leave your account.
  • Never trains AI on your data. Your email, calendar, and Slack are not used to train any model. Nerve uses Anthropic and AWS Bedrock commercial APIs, both contractually no-training, and we have not opted into any program that would change that. Providers auto-delete API inputs and outputs within 30 days.
  • Never sells or shares your personal data with advertisers or data brokers.

Where your data lives

  • Encrypted in transit (TLS) and at rest. OAuth tokens are encrypted with AES-256-GCM before storage.
  • Stored in the United States (Upstash Redis for working state, Neon Postgres for long-term metrics), with per-user and per-org isolation enforced at the storage layer.
  • Raw email bodies are not persisted after analysis. We keep the summarized output and the metadata needed for follow-up. Briefings roll off on a 90-day window.
  • Processed by Anthropic and AWS Bedrock for AI. Both are SOC 2, as are Clerk (auth), Upstash, Vercel, and Stripe (we never see your card; Stripe is PCI DSS Level 1).

Deleting your data

Delete your account anytime from your account settings or by emailing patrick@getnerve.ai. We remove your data and revoke Nerve's access to your connected accounts within 30 days.

The honest caveats

We request broad email and calendar permissions today. Because the agents execute actions you approve, the scopes include send and edit, not just read. The control is that nothing executes without your specific approval. Incremental, read-only-first scopes are how we are fixing the up-front ask.

We are not SOC 2 certified yet. We are a solo-founder company built in under 90 days. Everything we run on is SOC 2 (Clerk, Upstash, Vercel, Anthropic, Stripe), but Nerve itself is not certified, and we will not pretend otherwise. Business customers can request a DPA at patrick@getnerve.ai.

Our marketing site uses analytics and B2B visitor identification. PostHog (Do Not Track respected) for product analytics, and Apollo for company-level visitor identification on US business traffic. This runs on our public pages, not inside your authenticated workspace, and never touches your email or calendar data. Detail and opt-out in the Privacy Policy.

We are US-first. Data is stored in the US, and we do not yet offer EU data residency or a full GDPR consent flow. If you are an EU-regulated buyer, talk to us before you connect.

Report a vulnerability

Found a security issue? Email security@getnerve.ai. Our disclosure policy is at /.well-known/security.txt. Please give us a chance to fix it before posting publicly.

Read the detail

Privacy Policy  ·  Security  ·  Terms

Last updated 2026-06-07.